rdf - Type inference using class expresions on Protege

Question

I was hoping I could defined Elective_Course in such a way that all individuals that are partOf a Knowledge_Unit and are not of type Core_Course are Elective_Courses.

For Course I have this:

I thought maybe something on the lines of (partOf only Knowledge_unit) and not(Core_Course) would do the trick, but it doesn't work.

Maybe it is because I need and only with the inverse property of partOf, but I haven't defined that property since I didn't need it for the rest. So I could say in Knowledge_Unit that hasPart only Course and then not(Core_Course) on the Elective_Course.

But I'm lost, so any help or links to examples are welcome.

The full ontology is available here

Answer 1

The problem you are experiencing is due to the open world assumption in OWL and Description Logics. This means the reasoner will only infer new inferences from explicitly stated information and not from absence of information.

So to achieve your desired result you have to state that Core_Course and Elective_Course are disjoint. Then you can do one of two things:

(1) State for each individual that is an elective that it is of type not (Core_Course). But then you could just as well state it as of type Elective_Course.

(2) Define all your knowledge unit individuals as say core1, core2, ... coren, elective1, elective2, ..., electivek. Then set Knowledge_Unit equivalent to all individuals and Core_Course to all core course individuals, i.e.

 Class: Knowledge_Unit
   EquivalentTo: {core1, core2, ... coren, elective1, elective2, ..., electivek}
 Class: Core_Course
    SubClassOf: Knowledge_Unit,
    EquivalentTo: {core1, core2, ... coren}

Update: 20180510

Wrt universal property restrictions like partOf only KnowledgeUnit, it is important to understand that it also includes individuals for which it is known that they do not take part in any partOf relation. This can lead to entailments that can be difficult to understand. I have written about this here.